Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:120481
Content-Type: text/plain;
	charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3731.600.7\))
In-Reply-To: <CAB-BujZpcCo8kmvHDheOd_ZxSORCiqHj-o1XSYJFB0VDn0WM1Q@mail.gmail.com>
Date: Wed, 31 May 2023 02:09:36 +0200
Cc: php internals <internals@lists.php.net>
Content-Transfer-Encoding: quoted-printable
Message-ID: <5140E8C4-A27B-40C0-82F2-A73700A3D3A0@gmail.com>
References: <289E585B-EF8B-4B17-89BE-BE8295FD9FE1@gmail.com>
 <CAB-BujYyz4vFAmi+OKi0rjqViE02eUjDpkkYyAMRCf10t8mNXA@mail.gmail.com>
 <4CA1668E-A342-452E-A994-5839C377CB27@gmail.com>
 <CAB-BujY2f3tM5CwRVAuVj962iC_9K=1X1A0at2QVdORxQDqx3A@mail.gmail.com>
 <B35AAF0F-B899-4771-AAFD-1109BE7A9B7D@gmail.com>
 <CAB-BujYgoMVBgTYfgK+iQKyFB0HGGjQkdyV6e8o8=1KwV5mdzw@mail.gmail.com>
 <9D69B8D4-835A-47C0-A3BD-64285849BB02@gmail.com>
 <CAB-BujZpcCo8kmvHDheOd_ZxSORCiqHj-o1XSYJFB0VDn0WM1Q@mail.gmail.com>
To: Andreas Hennings <andreas@dqxtech.net>
Subject: Re: [PHP-DEV] [RFC] [Discussion] Add new function `array_group`
From: buritomath@gmail.com (Boro Sitnikovski)


> On 30.5.2023, at 21:56, Andreas Hennings <andreas@dqxtech.net> wrote:
>=20
> On Tue, 30 May 2023 at 19:25, Boro Sitnikovski <buritomath@gmail.com> =
wrote:
>>=20
>> Here's some more examples:
>>=20
>> 1. Use `array_group` to create list of singleton list:
>> ```
>> $groups =3D array_group( $arr, function( $p1, $p2 ) {
>>  return false;
>> } );
>> ```
>>=20
>> (This can also be achieved with `array_map` returning `[ $x ]`)
>>=20
>> 2. Distinct groups for consecutive positive and negative elements
>>=20
>> ```
>> $arr =3D [-1,2,-3,-4,2,1,2,-3,1,1,2];
>>=20
>> $groups =3D array_group( $arr, function( $p1, $p2 ) {
>>  return ($p1 > 0) =3D=3D ($p2 > 0);
>> } );
>> ```
>>=20
>> This produces `[[-1],[2],[-3,-4],[2,1,2],[-3],[1,1,2]]`, so we can =
easily capture the groups of highs/lows for example.
>>=20
>> 3. Group sentences (similar to `explode`, but still different)
>>=20
>> ```
>> $arr =3D "Hello, PHP. Good to see you.";
>> $groups =3D array_group( str_split( $arr ), function( $p1, $p2 ) {
>>        return '.' !=3D=3D $p1;
>> } );
>>=20
>> $groups =3D array_map( 'join', $groups );
>> ```
>>=20
>> Producing `[ "Hello, PHP.", " Good to see you." ]`.
>>=20
>> 4. Grouping book sections
>> ```
>> $book_sections =3D [ '1.0', '1.1', '1.2', '2.0', '2.1', '3.0', '3.1' =
];
>> $groups =3D array_group( $book_sections, function( $p1, $p2 ) {
>>        return $p1[0] =3D=3D=3D $p2[0];
>> } );
>> ```
>>=20
>> Producing `[ [ '1.0', '1.1', '1.2' ], [ '2.0', '2.1'], [ '3.0', '3.1' =
] ]`
>>=20
>> and so on...
>>=20
>> Basically, it's a very general utility :)
>>=20
>> Best,
>>=20
>> Boro
>=20
> Only one of these examples actually uses a comparison function that
> actually compares two adjacent values.
> All the others look at a single value only, or they don't care about
> the original order at all.
>=20
> And that example (2) seems quite artificial and rare.
>=20
>> so we can easily capture the groups of highs/lows for example
>=20
> How often does this problem come up in practice?
>=20
> I did actually have a look at the Haskell doc page you linked.
> =
https://hackage.haskell.org/package/groupBy-0.1.0.0/docs/Data-List-GroupBy=
.html
> I wonder where this is used in the Haskell world, and how often they
> need this true comparison of adjacent values.
> The examples from the doc page are quite artificial. That's ok for a
> doc page, but does not make a good argument for why we need it.

Thanks for digging deeper in trying to understand the value that this =
would bring.

For a more realistic example, check the following gist: =
https://gist.github.com/bor0/b5f449bfe85440d96abd933b9f03b310#file-test_da=
tetime_group-php. This was a very recent thing I worked on in =
production. The events came ordered from the database. But also, this is =
not the first time I was in need of a grouping functionality.

Sorry for repeating, but I see it as a handy utility. We can bring the =
same argument for any of the `array_*` functions because most of them =
can be implemented with for loops.

If the main concern is the consecutive processing of the elements, we =
could add a third argument to `array_group` (flag) that would allow =
users to alter the processing behaviour - at a price, consecutive taking =
O(n) vs all elements taking something close to O(n^2)=